Accuracy in Metaphysics Tools: What Can Be Checked?

Accuracy in a metaphysics tool, as far as you can check it, means the engine repeats. The same birth data returns the same BaZi pillars. The same almanac window returns the same order. What you cannot check is whether a life will match a hope. BaZi and Pick a Good Day are professional guidance. Repeatable labels are not a promise of fate, and we do not call that guidance entertainment.

Key facts

Pillars are a closed calculation from a clock and a calendar rule, including Start of Spring and solar terms. Re-entering the birth should not invent a new day master.
Luck steps appear only with an hour and sex at birth. Their absence when the hour is blank is a correct result, not an inaccuracy.
A locked travel window, 1–21 October 2026, with no birth date, ranks 13 October first in this engine. That check is about order, not about anyone’s trip.
A sample birth of 2 March 1991 changes a travel score on 6 October 2026 because of clash penalties. The change should reappear if you enter the same pair again.
Western transits are not in this engine, so we cannot be “accurate” about them. We can only be clear that we do not compute them.
No study, rating, or client story is offered here as proof of life outcomes. We will not invent one.

The layer you can audit

Auditing starts with a pencil. Write the inputs. Write the outputs. Submit again. For BaZi, the outputs to compare are the pillars you can see, the day master, and whether steps are present or honestly absent. The product that prints them is BaZi, with the longer layer at the full chart. The rules of the clock are in what BaZi is and four pillars. If two runs on the same inputs disagree, that is a defect. Tell Help. Do not paper over it with interpretation.

For the almanac, the outputs are the ordered days, the scores, and the why-text: officer, path, activity match, clash notes. Pick a Good Day records the public weights. The tool is Pick a Good Day. The October 2026 travel check is a public example because a vague claim of accuracy is useless. Thirteen October, in that window, with no birth date, is the top row we lock. If you need a different occasion, you need a different run. Accuracy does not mean every occasion crowns the same Tuesday.

The 1991-03-02 sample shows the personal layer moving a number for a reason you can name: the candidate day clashes the year branch in force after Start of Spring. If a tool moved the number and could not say why, you could not audit it. Ours is built so the penalty has a size. Size is not a probability. A repeatable penalty can still sit on a day you should keep because of a lease.

A reader who wants the trip to prove the score

Someone books 13 October 2026 because it topped a travel window, has a miserable connection, and concludes the engine is false. Someone else has a pleasant weekend and concludes it is true. Neither trip is the test. The test is whether 13 October still tops that same window when the birth field is empty. Weather, crews, and expectations are not fields in the score. The disclaimer says the rank is guidance. Accuracy talk that forgets that sentence turns a calendar check into a superstition about delays.

If they add the sample birth or their own and the order moves, the check becomes: did it move for the clash we document, and does it move the same way tomorrow? If yes, the personal layer is doing the only thing it claims. It is not doing fortune-telling quality control. We do not have a second, secret score for “how the day will feel.”

What accuracy language usually smuggles

“More accurate than Western astrology” smuggles a contest we will not run. The systems print different objects. You can check each for internal repeatability. You cannot crown a winner with a shared life score without inventing the score. “100% accurate” smuggles a guarantee about events. Refuse that sentence in both traditions. A practitioner who offers it is offering something a calendar cannot pay.

Historical agreement is another smuggle. A pillar that matches a classical worked example is evidence the calendar rule was followed. It is not evidence that a modern hiring decision should follow the pillar. Keep the audit on the symbols. Keep the decision on the decision’s own facts: the contract, the clinic, the rota.

Children’s charts are not a test set. Do not enter a minor’s birth to see if the engine “works.” Adults only. The past-life quiz is not an accuracy product either. It is fiction. An imagined story cannot be checked against a registry of past lives, because we are not claiming a registry.

A checklist that fits on a card

Same inputs, same pillars. Blank hour, blank hour pillar, no steps. Same window, same order, and the locked October example still holds for that example. Clash penalty appears for a reason you can point to. No percent chance of success appears. No medical date appears. If a checkbox fails, the tool has drifted. If every checkbox holds and your week is still hard, the tool has not failed its check. Your week was never the unit.

Use the check, then decide in the world

Audit once when you are calm, not every morning of a bad week. A stable engine will not soothe a late train, and it should not be asked to. Plans for a longer day list are described on the plans page. They do not buy a truer life. They buy more rows of the same method only after a plan can be offered. It cannot be offered now.

What this is not

A clinical trial, a numbered grade, or a percentage of predictions hit.
Proof that a top-ranked day will feel good.
A Western ephemeris check. We do not run that sky.
Permission to test the engine on a child’s data.
A claim that repeatability equals destiny.
A purchase. Opening the tool does not charge you.